Weird identifier problem in TMLanguage syntax highlighting for AS3

[hydroperfox]

If anyone here is familiar with TMLanguage for Visual Studio Code, I appreciate any help. I've finished a basic TMLanguage grammar for AS3 only, but I'm getting a weird highlighting with the snippet:

reviver = JSBridge.toJSFunction("");

For some reason, it does not highlight the string literal inside the call. But, if I do

reviver = Aaa.aaaaaaaaaaaaaaaaa("");

It works. Comparison:

It does not happen only with string literal; it happens with function expression and anything else.

Highlights of my grammar:

# repository: directive (merges directives + statements + expressions)

# Parenthesized
- begin: \(
        end: \)
        name: expression.group
        patterns:
          - include: '#directive'

# String literal
- begin: \@"""
  end: \"""
  name: string.quoted.triple
- begin: \@'''
  end: \'''
  name: string.quoted.triple
- begin: \"""
  end: \"""
  name: string.quoted.triple
  patterns:
    - include: '#stringLiteral'
- begin: \'''
  end: \'''
  name: string.quoted.triple
  patterns:
    - include: '#stringLiteral'
- begin: \@"
  end: \"
  name: string.quoted.double
- begin: \@'
  end: \'
  name: string.quoted.single
- begin: \"
  end: \"
  name: string.quoted.double
  patterns:
    - include: '#stringLiteral'
- begin: \'
  end: \'
  name: string.quoted.single
  patterns:
    - include: '#stringLiteral'

# Dot handling
- match: (\.)\s*({{id}})
  captures:
    1: { name: punctuation }
- begin: (\.)\s*(\<)
  end: \>
  patterns:
    - include: '#directive'
- match: \.\.
  name: keyword.operator
- match: \.
  name: punctuation

Here's the full grammar

[ajwfrost]

Not sure what's going on there but have you tried looking at the AS3 compiler, lexical analyser etc?
https://github.com/apache/royale-compiler/blob/develop/compiler/src/main/jflex/org/apache/royale/compiler/internal/parsing/as/RawASTokenizer.lex
is the grammar, the royale-compiler stuff is all pretty similar to the Falcon compiler provided in the AIR SDK.

[hydroperfox]

I have these in Rust, and I'd provide semantic tokens later in the VSCode extension.

Particularly though, VSCode language support starts with a TextMate language (TMLanguage) grammar, which is then overriden by a semantic/tokenizer later.